Earticle

본 연구는 실시간 객체 탐지 기술인 YOLO에 대해 YOLOv8부터 YOLO12까지의 버전과 모델에 대해 소규모 커스텀 데 이터셋 환경에서 성능평가를 수행하여, 제한된 데이터와 자원을 보유한 실제 응용 환경에서의 모델 선택 가이드라인을 제시한 다. 교통 환경 기반 6개 클래스, 2,664장, 총 6,354개의 라벨링 커스텀 데이터셋을 구축하고, Google Colaboratory(NVIDIA Tesla T4 GPU) 환경에서 동일한 학습 조건으로 YOLOv8부터 YOLO12까지 총 26개 모델을 학습·평가하였다. 연구 결과, YOLOv9m이 mAP50-95 0.520으로 자체평가 기준 중 최고 성능을 기록하였으며, 대부분의 버전에서 소형(s) 또는 중형(m) 모 델이 시간 대비 뛰어난 성능을 보였다. 대형(l) 및 초대형(x) 모델은 학습 시간이 2~3배 증가했음에도 성능은 정체되거나 감소하 였다. 공식 벤치마크와 달리 소규모 데이터셋에서는 중간 크기 이상 모델에서 성능이 감소하는 비선형적 패턴이 나타났으며, YOLO 버전보다 모델 크기 선택이 성능에 더 큰 영향을 미치는 것으로 확인되었다. 본 연구는 제한된 데이터 환경에서의 실용적 버전 및 모델 선택 기준을 제시하며, 향후 하이퍼파라미터 최적화 및 도메인이나 데이터 규모에 따른 비교 연구가 필요하다.

This study evaluates real-time object detection models in the YOLO series, from YOLOv8 to YOLO12, using a small-scale custom dataset and proposes practical guidelines for model selection in resource-constrained environments. A custom traffic-based dataset consisting of 2,664 images, 6 classes, and a total of 6,354 labeled objects was constructed. All experiments were conducted under identical training conditions using Google Colaboratory with an NVIDIA Tesla T4 GPU, and a total of 26 models from YOLOv8 to YOLO12 were trained and evaluated. The results show that YOLOv9m achieved the highest performance among the evaluated models, with an mAP50–95 of 0.520. In most versions, small (s) and medium (m) models demonstrated superior performance relative to training time. Large (l) and extra-large (x) models showed required 2–3 times longer training, while their performance remained stagnant or decreased. Unlike official benchmarks, a nonlinear performance pattern was observed in the small-scale dataset, where performance decreased in models of medium size or larger. Furthermore, model size had a greater impact on performance than the YOLO version itself. This study provides practical guidelines for selecting appropriate YOLO versions and model sizes in resource-constrained environments. Future work should explore hyperparameter optimization and extend comparative analyses across different domains and dataset scales.

최근 생성형 인공지능의 발전은 학술 콘텐츠 생성 방식에 큰 변화를 가져오고 있으며, 특히 논문 작성 자동화에 대한 수요가 증가하고 있다. 본 연구는 논문작성 템플릿을 기반으로 생성형 인공지능(ChatGPT, Gemini, Copilot) 을 활용하여 자동 논문 생성 시스템을 설계하고 성능을 비교하였다. 연구는 템플릿 기반 구조를 정의하고, 각 AI 모델에 동일한 프롬프트를 입력하여 결과물을 도출한 후 속도, 정확도, 인식율(F1 Score), 사용자 평점 등 다양한 지표로 성능을 분석하였다. 실험 결과, 각 모델은 고유한 특성과 성능 차이를 보였으며, 템플릿 기반 접근 방식은 일관된 구조와 품질 향상에 기여하였다. 본 연구는 학술 콘텐츠 자동화의 효율성을 높이고, 연구자들이 보다 창의적인 작업에 집중할 수 있는 기반을 제공할 수 있다. 향후 다양한 분야에 적용 가능한 템플릿 확장 및 사용자 맞춤형 생성 기능 개발이 기대된다.

Recently, the advancement of generative AI has significantly transformed academic content creation, especially increasing the demand for automated paper writing. This study designs a template-based automatic paper generation system using generative AI models (ChatGPT, Gemini, Copilot) and compares their performance. The research defines a structured template, inputs identical prompts into each AI model, and analyzes the results based on speed, accuracy, F1 Score, and user ratings. Experimental results show distinct characteristics and performance differences among the models, and the template-based approach contributes to consistent structure and improved quality. This study enhances the efficiency of academic content automation and provides a foundation for researchers to focus on more creative tasks. Future work includes expanding templates for various fields and developing personalized generation features.

본 논문에서는 학령인구 감소에 따른 대학 중도탈락 문제의 중요성을 인식하고, 중도탈락 학생 예측을 위한 다양한 머신러닝 알고리즘 의 성능을 비교 분석하였다. 데이터의 수집은 H 대학 신입생 3,435명의 데이터를 활용하였으며, 신입생 입학 후 1학기 축적된 데이터 13개를 입력 특성으로 사용하였다. 이 데이터를 이용하여 KNN, Logistic Regression, Decision Tree, Random Forest, Gradient Boosting, SVM, Neural Network 총 7가지 이진 분류 모델을 실제 대학 신입생 데이터를 활용하여 평가하였다. 실험 결과 Random Forest 모델이 가장 우수한 예측 성능을 보였으며, Decision Tree와 SVM이 뒤를 이었다. 반면, Logistic Regression과 Gradient Boosting은 상대적으로 낮은 성능을 나타냈다. 이 연구는 대학 중도탈락 예측을 위한 최적의 모델 선택에 기여하고, 향후 추가 연구 방 향을 제시한다.

This study recognizes the importance of addressing the issue of college dropout due to the decline in the population of students with higher education and compares the performance of various machine learning algorithms for predicting student dropout. The data used in this study consists of information from 3,435 new students enrolled at H University, with 13 accumulated data features collected during their first semester. Using this data, seven binary classification models—K-Nearest Neighbors (KNN), Logistic Regression, Decision Tree, Random Forest, Gradient Boosting, Support Vector Machine (SVM), and Neural Network—were evaluated in terms of their predictive performance. The experimental results show that the Random Forest model outperformed the other models in terms of prediction accuracy, followed by Decision Tree and SVM. In contrast, Logistic Regression and Gradient Boosting demonstrated relatively lower performance. This research contributes to the selection of the optimal model for predicting college dropout and suggests potential directions for future studies.

ESG 경영의 중요성이 커지면서 단순 재무 성과만으로는 기업의 지속 가능성을 평가하는 데 한계가 있으며, 기업의 지속 가능성을 평가하는 새로운 방법들이 요구되고 있다. 본 연구는 기업의 재무 데이터를 활용하여 ESG 등급(E, S, G)을 예측하는 머신러닝 모델을 개발하고, 이를 통해 ESG 등급의 예측 가능성을 검토하는 데 목적을 두었다. 2021년 한국거래소에 상장된 기업들을 대상으로 데이터 수집이 이루어졌으며, 랜덤 포레스트, 서포트 벡터 머신, 로지스틱 회귀 등 다양한 머신러닝 알고리즘을 적용하여 ESG 등급을 예측하였다. 분석 결과, 랜덤 포레스트 모델이 E 및 S 등급에서 가장 높은 예측 정확도를 보였으며, G 등급에서는 상대적으로 낮은 성능을 나타냈다. 이를 통해 ESG 등급의 예측에 재무 데이터가 중요한 역할을 할 수 있음을 확인하였다. 본 연구는 기업과 투자자 모두에게 중요한 의사결정 도구를 제시함으로써 ESG 경영의 실질적 강화에 기여할 것이다.

The purpose of this study was to compare marbling score, meat quality, juiciness, sarcomere length, and skeletal muscle satellite cell (SMSC) growth and related gene expression between Woori black pig (WB) and the Landrace, Yorkshire, and Duroc (LYD) crossbreed at different body weights (b.w.). WB was developed to improve meat quality and growth efficiency by crossbreeding Duroc with Korean native black pig. A total of 24 pigs were sacrificed when their b.w. reached about 50, 75, 100, and 120 kg. SMSC were isolated from the femoris muscles, and muscle and adipose tissues were sampled from the middle and the subcutaneous part of the femoris of hind legs, respectively. Expression levels of genes including Myoblast determination protein 1 (MyoD), Paired box gene 3 (Pax3), Myosin heavy chain (MyHC), and Myogenin, which are responsible for the growth and development of SMSC, were higher in LYD than the WB. Muscle growth inhibitor myostatin (MSTN), however, was expressed more in WB compared to LYD (p < 0.01). Numbers of SMSC extracted from femoris muscle of LYD at 50, 75, 100, and 120 kg b.w. were 8.5 ± 0.223, 8.6 ± 0.245, 7.2 ± 0.249, and 10.9 ± 0.795, and those from WB were 6.2 ± 0.32, 6.2 ± 0.374, 5.3 ± 0.423, and 17.1 ± 0.315, respectively. Expression of adipogenic genes in adipose tissue including CCAAT/enhancer-binding protein (CEBP)-β, peroxisome proliferator activated receptor (PPAR)-γ, and fatty acid synthase (FASN), were greater in WB when compared with LYD (p < 0.01). Results from the current study suggest that different muscle cell numbers between 2 different breeds might be affected by related gene expression and this warrants further investigation on other growth factors regulating animal growth and development.

Background Logistics service workers face elevated injury rates, with ankle-related incidents comprising 23% of workplace injuries. Traditional binary foot posture classifications may oversimplify the complex biomechanical relationships affecting dynamic balance performance in occupational settings. Purpose To apply unsupervised machine learning clustering to identify distinct foot posture phenotypes among logistics service workers and compare dynamic balance performance between identified clusters using the Y-Balance Test. Study design Cross-sectional observational study Methods A total of 112 logistics service workers were analyzed using K-means clustering based on age, body mass index, work duration, navicular drop, and resting calcaneal stance position. Silhouette score analysis determined optimal cluster number. Dynamic balance was assessed using the Y-Balance Test, measuring reach distances in anterior, posteromedial, and posterolateral directions. Results Four distinct phenotypes emerged: rearfoot valgus-dominant pronated (n=22), midfoot collapse-dominant pronated (n=32), age-related (n=23), and supinated (n=35) foot types. Significant differences in dynamic balance performance were observed in the posterolateral direction (F=3.900, p=0.011). The supinated phenotype demonstrated superior posterolateral reach performance (77.76±13.63%) compared to midfoot collapse-dominant (66.18±16.37%, p=0.003) and age-related phenotypes (67.78±14.97%, p=0.018). Conclusions Unsupervised machine learning successfully identified naturally occurring foot posture phenotypes with distinct dynamic balance characteristics. Midfoot collapse-dominant pronation demonstrated greater balance impairments than rearfoot valgus patterns, supporting the implementation of phenotype-specific interventions for workplace injury prevention in logistics workers.

Purpose: This study aimed to explore the differences between patient and nurse perceptions of nursing needs and performance in integrated nursing care service wards within small to medium-sized hospitals. Methods: The participants of the study were 116 patients and 116 nurses in the integrated nursing care service wards of small and medium-sized hospitals in U and Y cities. Data collection was conducted between July 18 and August 12, 2022. The collected data were analyzed using descriptive statistics, the paired t-test, the independent t-test, and the importance-performance analysis. Results: Only the difference between nurse-perceived nursing needs and nursing performance (gap 2) was statistically significant (t=4.18, p<.001). Patients' physical needs were higher than what nurses perceived, and patients' financial were higher than nursing performance as perceived by both nurses and patients. In addition, nurses' performance was lower than the perceived nursing needs in all domains. According to the importance-performance analysis, the "concentrate here" areas included infection control, respect for patients' rights, and education of patients and families about diseases and treatment plans. Additionally, financial issues in nursing were identified as a "low priority" area requiring long-term intervention. Conclusion: It is crucial to prioritize nursing care that addresses patients' needs for infection control, respect for rights, and education. Nurses should also meet the physical and economic support needs of patients. Furthermore, it is recommended to adjust the nurse staffing ratio and foster systematic collaboration among hospitals, taking into account the circumstances of small to medium-sized hospitals.

Exercise plays an important role in regulating energy homeostasis, which affects the diversity of the intestinal microbial community in humans and animals. To the best of the authors’ knowledge, few studies have reported the associations between horse gut microbiota along with their predicted metabolic activities and the athletic ability of Jeju horses and Thoroughbreds living in Korea. This study was conducted to investigate the association between the gut microbiota and athletic performance in horses. This study sequenced the V3 and V4 hypervariable regions of the partial 16S rRNA genes obtained from racehorse fecal samples and compared the fecal microbiota between high- and low-performance Jeju horses and Thoroughbreds. Forty-nine fecal samples were divided into four groups: high-performance Jeju horses (HJ, n = 13), low-performance Jeju horses (LJ, n = 17), high-performance Thoroughbreds (HT, n = 9), and low-performance Thoroughbreds (LT, n = 10). The high-performance horse groups had a higher diversity of the bacterial community than the low-performance horse groups. Two common functional metabolic activities of the hindgut microbiota (i.e., tryptophan and succinate syntheses) were observed between the low-performance horse groups, indicating dysbiosis of gut microbiota and fatigue from exercise. On the other hand, high-performance horse groups showed enriched production of polyamines, butyrate, and vitamin K. The racing performance may be associated with the composition of the intestinal microbiota of Jeju horses and Thoroughbreds in Korea.

Objective: The purpose of this study was to investigate whether there is a difference in the brain connectivity in mental practice and physical performance of training bilateral upper extremity function. Method: The subject performed activities involving mental tasks and physical exercise for bilateral upper extremity functioning during each phase of EEG measurements. The subject performed a symmetrical task(lifting a box and placing it back) that involved moving both arms at the same time and an asymmetrical task(opening and closing a bottle cap) in order to perform functional tasks. EEG electrodes were attached to Fp1, Fp2, F3, F4, T3, T4, P3, and P4. Data analysis was performed using Cross-Line Mapping for correlational analyses between EEG electrode pairs. Conclusion: This study found that the brain connectivity patterns of symmetrical and asymmetric upper extremity tasks have similar patterns for the motor and sensory area, and that the correlation of the physical practice is generally higher than that of the mental practice.

Two experiments were conducted to compare the effects of different organic acids on growth performance and apparent nutrients digestibility in weaned pigs. In both the experiments, 180 pigs were assigned to four treatments with three replicates comprising of 15 pigs in each. Formic acid, ammonium-formate, lactic acid, and acid mixture were added to diets at 0.50%(Exp. 1) and 0.30%(Exp. 2) as dietary treatments for 5 and 6 wk feeding trial, respectively. The acid mixture was prepared by mixing formic acid and lactic acid at 50:50 ratios. To investigate the apparent ileal amino acids digestibility, twelve pigs(3 per treatment) were used and fitted with simple ileo-caecal T-cannula for both experiments. In Exp.1, growth performance was com- parable(P>0.05) among pigs fed different organic acids, while acid mixture had higher(P<0.05) weight gain than that of lactic acid in Exp.2. The apparent ileal digestibility of amino acids was highest(P<0.05) in pigs fed acid mixture and lowest(P<0.05) in pigs fed formic acid diets in both experiments. These results indicated that supplementation with acid mixture(formic acid and lactic acid) improved performance and ileal amino acid digestibility in weaned pigs.

In this research, the growth efficiency, nutritional utilization, fecal microbial levels, and fecal score of weaned pigs were evaluated using therapeutic zinc oxide (ZnO) and zinc aspartic acid chelate (Zn-Asp). In a 42-day feeding trial, 60 weaned pigs ([Yorkshire × Landrace] × Duroc) were arbitrarily allotted (age: 21 days; 7.01 ± 0.65 kg preliminary body weight) to 3 different treatment groups with 5 repetitions (2 male and 2 female piglets) in each pen. The trial had 2 different phases, including 1?21 days as phase 1, and 22?42 days as phase 2. The nutritional treatments were: basal diet as control (CON), basal diet incorporated with 3,000 ppm ZnO as TRT1, and basal diet incorporated with 750 ppm Zn-Asp as TRT2. In comparison to the CON group, the pigs in the TRT1 and TRT2 groups had greater (p < 0.05) body weight on day 42; an average daily gain, and an average daily feed intake on days 22?42. Furthermore, during days 1-42, the average daily gain in the treatment groups trended higher (p < 0.05) than in the CON group. Additionally, the fecal score decreased (p < 0.05) at week 6, the lactic acid bacteria count tended to increase (p < 0.05), and coliform bacteria presented a trend in reduction (p < 0.05) in the TRT1 and TRT2 groups compared to the CON group. However, there was no difference in nutrient utilization (p > 0.05) among the dietary treatments. Briefly, the therapeutic ZnO and Zn-Asp nutritional approaches could decrease fecal score and coliform bacteria, increase lactic acid bacteria, and improve growth efficiency; moreover, Zn- Asp (750 ppm) can perform a comparable role to therapeutic ZnO (3,000 ppm). So we can use Zn-Asp (750 ppm) instead of therapeutic ZnO (3,000 ppm) for the better performance of weaning pigs and the reduction of environmental pollution, as therapeutic ZnO is responsible for environmental pollution.

본 논문에서는 주성분 회귀법과 부분최소자승 회귀법을 비교하여 보여준다. 이 비교의 목적은 선형형태를 보유한 근적외선 분광 데이터의 분석에 사용할 수 있는 적합한 예측 방법을 찾기 위해서이다. 두 가지 데이터 마이닝 방법 론인 주성분 회귀법과 부분최소자승 회귀법이 비교되어 질 것이다. 본 논문에서는 부분최소자승 회귀법은 주성분 회귀법과 비교했을 때 약간 나은 예측능력을 가진 결과를 보여준다. 주성분 회귀법에서 50개의 주성분이 모델을 생 성하기 위해서 사용지만 부분최소자승 회귀법에서는 12개의 잠재요소가 사용되었다. 평균제곱오차가 예측능력을 측 정하는 도구로 사용되었다. 본 논문의 근적외선 분광데이터 분석에 따르면 부분최소자승회귀법이 선형경향을 가진 데이터의 예측에 가장 적합한 모델로 판명되었다.

빠르게 진화하는 정보 통신 기술로 인해서 새로 생성되는 데이터는 급속히 증가하고 있다. 센서, 스마트 폰, 소셜 네트워킹 사이트, 논리적 데이터 및 이니셔티브의 데이터가 이 엄청난 정보 폭발을 발생시키고 있다. Hadoop은 방대한 양의 데이터를 처리하는 데 사용되는 프레임워크이며 기업에서 중요한 정보를 저장하고 관리하 는데 많이 사용되고 있다. 또한, Hadoop은 분산형 파일 시스템으로 기능 및 보안 업데이트가 계속 진행되고 있다. 그러나 Hadoop은 민감한 데이터의 보안에 취약하다. 본 연구에서는 클라우드 시스템에 대한 보안 문제를 조사하 였다. 특히, 현재 클라우드 환경에서 사용할 수 있는 암호화 알고리즘을 조사하고 데이터 암호화 실험을 수행하였 다. 실험에서는 Salsa20, 3-DES, AES, MARS, Blowfish 등 총 6가지 알고리즘을 비교 하였고 논문에서 제안하 는 Salsa20을 사용하는 것이 최고 22%의 성능 향상이 있는 것을 확인하였다.

Data is rising tremendously in the rapidly evolving world of information and communications technology. Data from sensors, smartphones, social networking sites, logical data, and initiatives are fueling this enormous information explosion. Hadoop is a framework used to process vast amounts of data and is widely used by companies to store and manage important information. Additionally, Hadoop is a distributed file system, and feature and security updates are continuously being made. In this paper, encryption algorithms currently available in the cloud environment were investigated and data encryption experiments were conducted. In the experiment, a total of six algorithms, including Salsa20, 3-DES, AES, MARS, and Blowfish, were compared, and it was confirmed that using Salsa20 proposed in the paper improved performance by up to 22%.

The purpose of this study was to compare performance-related physi-cal fitness factors and anaerobic power between national and backup players to enhance the performance of wheelchair badminton players and to discover and foster new athletes. This study included 12 wheel-chair badminton players divided into two groups: national players (n=7) and backup players (n=5). Collected data included cardiorespiratory endurance, muscle endurance, muscle strength, power, agility, and an-aerobic power. Results indicated that the national players achieved a significantly higher level of muscle endurance, peak power, and mean power than the backup players. Although none of the parameters showed statistically significant differences, the performance levels of the national players were higher than those of the backup players. Be-cause the physical fitness aspects of wheelchair players vary accord-ing to their ability and are essential factors, individualized training pro-grams for enhancing performance and preventing injuries among wheelchair badminton players should be developed.

This study investigates the treatment performance of two different carrier types in a moving bed biofilm reactor (MBBR). Two different carrier types were employed: regular shape carrier (RSC) and irregular shape granular (ISG) carrier. The results demonstrate that the airflow design should be based on the specific gravity and shape of carrier. Both the surface area COD (SCOD) removal rate and the attached biomass increased at a linear rate as the loading rate. The study found that the attached biomass on the ISG carriers was lower than those on the RSC carriers. This is due to hydrodynamic shear and abrasion.

Purpose: It is necessary for a good baseball player to have various physical attributes including good muscle strength and sensitive proprioception. Childhood and adolescence are very important in developing these physical attributes, as they are suitable periods for acquiring basic physical strength and favorable physical attributes. Methods: Therefore, this study focused on the comparison of use of technology, physical characteristics, biomechanics, lifestyle, and game usage in children and adolescents. This study aimed to conduct research using basic data for more effective training and guidance. The included participants were 39 middle school students (3rd graders) and 42 elementary school students (6th graders) who participated in the KBO Next Level Training Camp. Results: To summarize the results of this study, there was no significant difference in game usage between young and adolescent players, and no evidence could be derived that suggested game habits affect physical characteristics, kinematics, or technical skills. Secondly, there was a significant difference in the physical abilities of young and adolescent players, regardless of gaming habits. Thirdly, there were no differences in batting speed, attack angle, or order of body movement between young and adolescent players. However, there were significant differences between the average peak hand speed, vertical bat angle, pitch release speed, spin rate, and striking distance (p=.001). Conclusion: According to the conclusions drawn from this study, there were no differences in physical ability, kinematics, or performance between young and adolescent players according to gaming habits.

하수 3차 처리수 내 잔존하는 pharmaceuticals and personal care products(PPCPs)를 분석하고 UV-AOP에 의한 제거 여부를 분 석하였다. UV-AOP 공정에 사용된 산화제는 과산화수소였다. 3차 처리 이후의 극미량의 농도를 고려하여 고상추출법을 이용하여 농축하였다. 254 nm 단일 파장이 나오는 램프와 185 nm와 254 nm가 나오는 램프를 이용하였다. UV 조사량은 반응조 체류시간 으로 조정하였으며 3단계로 구분하였다. 조사량의 범위는 701~1,683 mJ/cm2이었다. 평균 제거율은 조사량별로 경향을 보였으며 254 UV nm램프에서 59.4~77.5%, 185/254 nm 램프에서 61.6~76.0%였다. 평균제거율을 비교하면 185/254 nm UV 램프가 약간 우세하였으나 물질별로 비교하면 큰 차이를 보이지는 않았다. 물질별로도 254 nm UV 램프가 우세한 제거율을 보인 종이 8종, 185/254 nm UV 램프는 3종이었으나 육안으로 식별 가능한 차이를 보인 종은 각각 5종이었다. 높은 조사량에서 85.0% 이상의 제 거율을 보인 종은 공통적으로 Iohexol, Iopamidol, Iopromide, Sulfamethoxazole 4종이었으며 그 외 60.0% 이상의 제거율을 보인 종은 Acesulfame, Benzotriazole, Naproxen 3종이었다.

The pharmaceuticals and personal care products (PPCPs) remaining in the sewage tertiary treated water were analyzed and removed by UV-AOP treatment. The oxidizing agent used in the UV-AOP process was hydrogen peroxide. PPCPs analysis was performed using a solid phase extraction method considering the trace concentration after the third treatment. A single wavelength UV lamp of 254 nm and a double wavelength UV lamp of 185 nm and 254 nm were used. The UV irradiation dose was adjusted to the hydraulic retention time of the reactor. The dose range was 701 to 1,683 mJ/cm2 and divided into three levels. The average removal rate showed different tend by dose, and was 59.4 to 77.5% in the 254 nm UV lamp and 61.6 to 76.0% in the 185/254 nm UV lamp. When comparing the average removal rate, the 185/254 nm UV lamp was slightly superior, but there was no significant difference when comparing by each product. By species, 8 species showed a higher removal rate of 254 nm UV lamp and 3 species of 185/254 nm UV lamp, but 5 species showed differences that were discernible with the naked eye. Iohexol, Iopamidol, Iopromide, and Sulfamethoxazole were the most common species that showed removal rates of over 85.0% at high doses. Acesulfame, Benzotriazole, and Naproxen were the three species that showed removal rates of over 60.0%.